Runtime observation

Audit the request your app actually sends

The @savedyouatoken/sdk package observes fully assembled Anthropic and OpenAI requests in your own process: system prompt, tool schemas, model, and measured traffic. The real model call is untouched and the audit runs after its response.

Install

npm install @savedyouatoken/sdk

Anthropic

import Anthropic from '@anthropic-ai/sdk';
import { wrapAnthropic } from '@savedyouatoken/sdk';

const anthropic = wrapAnthropic(new Anthropic());

await anthropic.messages.create({
  model: 'claude-sonnet-5',
  system,
  tools,
  messages,
});

OpenAI

import OpenAI from 'openai';
import { wrapOpenAI } from '@savedyouatoken/sdk';

const openai = wrapOpenAI(new OpenAI());

await openai.responses.create({
  model: 'gpt-5.4',
  instructions,
  tools,
  input,
});

Reports mature with traffic

The first report for a request shape uses a provisional workload. After enough observations and elapsed time, the SDK emits an updated report based on measured request rate, output length, and cache behavior. You can override workload fields when you know them.

Safe in the request path

Auditing is deferred until after the provider responds. A capture, analysis, or sink failure never reaches application code, and no extra model call is made.

Private by default

Prompt and tool text stay in your process by default. The optional dashboard sink sends only a redacted report of counts, dollar figures, and static finding identifiers.

You choose the destination

Development defaults to console output; production defaults to silence. Opt in to console, file, callback, or dashboard sinks. The file and network forms contain prompt-free report data.

SDK or CLI?

Use this SDK when prompts are assembled dynamically inside a live application and you need to inspect the real outbound request. Use the CLI when prompts already exist as files or you want a pull request budget check. The packages are separate:@savedyouatoken/sdk for runtime integration andsavedyouatoken for the command line.

Read the complete SDK package documentation

Choose by what you need to do

These surfaces share the same audit engine, but they run at different points in your workflow.

Web analyser

Audit one prompt in your browser

Paste a prompt and get an immediate, private cost and waste report.

Open the analyser

Runtime SDK

Observe assembled production requests

Audit the system prompt and tools your application actually sends at runtime.

npm: @savedyouatoken/sdk

You are here

CLI

Audit files and enforce budgets in CI

Run deterministic audits over prompt files from a terminal or build pipeline.

npm: savedyouatoken

Run the CLI

Agent kit

Let a coding agent run the CLI

Give Claude Code, Cursor, or another coding agent instructions for invoking the CLI.

npm: savedyouatoken

Get the agent kit

Monitor

Planned

Track cost and regressions over time

Historical reporting and regression alerts for teams are planned, not yet available.

Planned